Whirr currently provides a Java API for running services. There is not much documentation available for it yet, so the best way to use it is to look at the Hadoop and ZooKeeper unit tests, which show how to start and stop clusters.
Building from source
Build Whirr from svn trunk by typing the following:
$ svn co http://svn.apache.org/repos/asf/incubator/whirr/trunk/ whirr-trunk $ cd whirr-trunk; mvn -DskipTests install
Running a Cluster
An example of how to run a Hadoop cluster with Whirr can be found at http://github.com/hammer/whirr-demo.